Count Data Models for Injury Data from the National Health Interview Survey (NHIS)

Logistic regression has been widely used in analyzing injury data from the National Health Interview Survey (NHIS). However, since its dependent variable is dichotomized to be either “1” (presence of an injury incident) or “0” (absence of an injury incident), logistic regression cannot provide sufficient information for studying the pattern of multiple injury incidents. In this study, several count data models are developed and compared using injury count data from 2006-2011 NHIS. The Zero-Inflated Negative Binomial (ZINB) model turns out to be the optimal count data model for our data. The inferences made from the ZINB regression model are compared with those from the logistic regression model. The results indicate that ZINB model can explore injury proneness and predict the mean number of injuries in the injury-prone population. These goals cannot be achieved by logistic regression although it might fit the dichotomized data well.

Name thesis:degree_name

Master of Science

Level thesis:degree_level

masters

Discipline thesis:degree_discipline

Biostatistics

Grantor dc:publisher

The Ohio State University

Year dc:date

2013

Author dc:creator

• Peng, Jin

Contributors dc:contributor

• Nagaraja, Haikady

• Biostatistics

Statement dc:rights

• unrestricted
• This thesis or dissertation is protected by copyright: all rights reserved. It may not be copied or redistributed beyond the terms of applicable copyright laws.

Language dc:language

English

Repository record dc:identifier

http://rave.ohiolink.edu/etdc/view?acc_num=osu1365780835

OAI identifier oai:identifier

oai:etd.ohiolink.edu:osu1365780835